Claude Sonnet 5 – A New Model for Daily Work
Anthropic is getting ready to launch Claude Sonnet 5. The model has shown up at delivery partners. This means the launch will happen in 5 to 7 days. Sonnet is the main model people use every day. It helps write code, texts, and build AI agents.
The new version has a bigger context window. It will hold about 1 to 2 million tokens. A token is a small piece of text the model reads. A bigger window means better understanding of long instructions. The model will also understand pictures and diagrams better.
Anthropic might use a new tokenizer. A tokenizer splits text into small parts. The new version may use 30 percent more tokens. This will make the prompt cost more. But the model will think and analyze images better. This is a big change from the old Sonnet.
First Tests and Examples
In tests, the model made an SVG image of a Nintendo Switch 2 console. SVG is a format for vector graphics. The model did this without a reference picture. The image looked like a real photo. This shows how well it handles design tasks.
Sonnet 5 is also very fast. It works five times faster than the older Opus model. It is also 75 percent cheaper. For one million input tokens you pay 3 dollars. For one million output tokens you pay 15 dollars. This is a good price for such quality.
Mythos 6 – An Even Stronger Model on the Way
News from Anthropic labs points to a new model. It is even better than Mythos 5. The name could be Mythos 6 or 5.1. The company has not shared official details yet. We know the model already exists and works.
The new model handles long thinking tasks better. It can plan and do big jobs. Coding and motivation skills are higher. Sources say this is a real step forward. But the model is already locked inside the company.
This shows that progress does not stop. Even when models are not available to the public. Computing power goes to training new versions. The company checks what safety steps are needed. Maybe we will see this model in the future.
What This Means for Users
New models will be better at hard tasks. For example, writing complex programs. Or having conversations with many steps. But for now, there is no public access. We can only wait for the official launch.
Anthropic may decide to release the model. Or keep it for internal testing. Either way, the company is moving forward. New versions will appear sooner or later.
GPT-4.6 Pro and a New Voice Model from OpenAI
OpenAI is getting ready to launch GPT-4.6 Pro. The model may come out on Thursday this week. This will be a big update. In tests, the model showed new skills. It creates websites and designs interfaces better.
In one test, the model made a first-person game. The game worked in a single HTML file. It had rooms, a bathroom, and a movement system. The project was 700 kilobytes in size. The time to generate it was 40 minutes. This shows how smart this model is.
On top of that, OpenAI is adding a new voice model. It is called GPT BDI 1. It works in real time. It can interrupt the conversation and fix mistakes. It is like an assistant that listens and talks naturally. The feature is already being added to the ChatGPT app.
Tests and Comparisons
The voice model has knowledge from August 2025. It can count words while talking. It can also correct you in the middle of a sentence. This is a big step from the old GPT-4 Omni model. The conversation is smoother and more natural.
Compared to Gemini, this model is less multimodal. But it handles conversation better. Users testing it say it is a big improvement. The assistant is becoming more human-like.
Sakana Fugu – A New Japanese AI System
A new Japanese lab called Sakana AI showed its system. It is named Fugu and Fugu Ultra. These are not normal language models. They are systems that manage other models. They decide which model to use for each task.
Sakana says Fugu Ultra matches Mythos 5 and Fable 5. But tests show something different. The system works well but is not as strong. In a Crossy Road game test, there were problems. The controls were reversed, the camera was shaky.
Still, Fugu Ultra is very fast. It finished the task in 22 minutes. The older Opus 4.8 Ultra needed 79 minutes. The cost was also much lower. Fugu cost 7 dollars, Opus cost 37 dollars. This is an important benefit for companies.
Who Is This System For?
The Sakana system can be useful for companies. It lets you combine many models into one. This saves time and money. The company AI w Biznesie could use such a solution. For example, to automate processes or marketing.
But remember the limits. Benchmarks do not always show real skills. Real-life tests are more important. Fugu is an interesting idea, but not a finished product.
Summary and Meaning for Business
This week brings many new models. Claude Sonnet 5 will make daily work easier. Mythos 6 shows the power of internal growth. GPT-4.6 Pro with a new voice will change AI talks. Sakana Fugu brings a new way of working.
For companies, this is a chance for better tools. Models are faster and cheaper. You can use them for automation. The company AI w Biznesie helps put such solutions in place. From marketing to data management.
It is good to follow these changes. New models come out faster and faster. Each has its own good and bad points. Choosing the right one depends on your needs. Testing is key before putting them to use.
The future of AI looks bright. Models are getting smarter. They help with work, learning, and fun. Stay up to date with the news. Use them in your company.
No responses yet